In Volume 5: Poetry of the Pre-Mongol Period

previous chapter: Manuscript Collections

¶ Persian poetry (shiʿr or naẓm) is distinguished from prose (nathr) by two indispensable formal components: end-rhyme (qāfiyah) and quantitative metre (ʿarūḍ or wazn), i.e. the regular succession of long and short syllables. Both of these features—together with others—are shared by Persian and Arabic poetry. Both, moreover, are conspicuously absent from all demonstrably pre-Islamic poetic works in Iranian languages. The conclusion, thus, virtually imposes itself that the Persian Muslims borrowed the principles of rhyme and of quantitative metre from Arabic.

Of course, the Iranians had poetry long before Islam. The oldest surviving works of Iranian verse are the Gāthās and Yashts of the Avesta. Although the language, the religious concepts and the poetic imagery of the Avesta are all close to those of the Vedas of ancient India, there is no trace in the Avesta of a quantitative metre of the sort which we find in the Vedas or in Greek poetry. The ancient Iranians evidently had a different system of versification from their Indian cousins. At present, the majority opinion among Iranologists¹ is that the poetic structure of the Gāthās is based exclusively on the number of syllables per line, though a minority view² (to which the present author inclines) holds that the Gāthās conform to a system of accentual metre; more precisely, I think it most likely that the Ahunauuaitī Gāthā (=Yasna 28–34) and the Vahištōišti Gāthā (=Yasna 53) consist of verses with a fixed number of stresses and a variable number of unstressed syllables, while the three other Gāthās (Yasna 43–46; 47–50; 51) have verses with a fixed number of syllables as well as a fixed number, and more-or-less fixed position of stresses. As for the Yashts, Geldner’s view³ that these are constructed of verses with a fixed number of ¶ syllables still has supporters,⁴ though Henning⁵ maintained, with what seem to me very strong—if somewhat elliptically formulated—arguments, that the verses of the Yashts have a fixed number of stressed syllables only, while the number of unstressed syllables varies. The same accentual principle underlies, according to the penetrating analyses of Henning⁶ and Boyce,⁷ pre-Islamic Middle-Persian and Parthian poetry, of which a good number of examples have come down to us both in Zoroastrian and in Manichaean script. The earlier suggestion by Benveniste⁸ that Middle-Persian poetry was based on mere syllable-counting required such a degree of tampering with the extant texts that it can effectively be disregarded, but also the occasionally expressed claim that it conformed with some kind of quantitative system has never been substantiated. Thus, Neo-Persian poetry, in basing itself on a system of a strictly regulated succession of long and short syllables, without any regard to the number or position of stresses, marks a clean break with Iranian pre-Islamic poetry.

With rhyme the situation is perhaps slightly more complicated. Although there is no hint of deliberate rhyme in the Avesta, the Manichaean Middle-Persian and Parthian hymns, or such manifestly pre-Islamic Zoroastrian poems as the Draxt ī Āsūrīg and the Ayādgār ī Zarērān,⁹ there are a small number of early Persian poems that do rhyme, but have no regular system of quantitative metre. Two poems of this sort have been preserved in Zoroastrian script in the manuscripts of the Pahlavi texts,¹⁰ a collection of miscellaneous religious and non-religious writings assembled, apparently, as a school-book for learners ¶ of the Middle-Persian language. One of these,¹¹ a poem rhyming consistently in -ān, bewails the Arab conquest of Iran and pines for deliverance; it thus clearly belongs to the Islamic period.¹² Another text,¹³ also rhyming in -ān, laments the impermanence of the world. It has an internal rhyme in its first verse (as is canonic in Arabic and Neo-Persian qaṣīdahs), as well as a sort of refrain (every second verse ends with the words andar gēhān). Although a precise dating of the latter text is not possible, there is no patent reason why it too should not belong to the Islamic period. In Arabic script we have the text which the anonymous Tārīkh i Sīstān¹⁴ says was inscribed on the fire-temple at Karkōy and which consists of eight periods rhyming in -ōsh followed by one unrhymed line. The verses, if authentic, are presumably transcribed from an original in Zoroastrian script. Of course, it is well known that Zoroastrians continued to compose Middle-Persian documents of various sorts well into the Islamic period. It is thus in principle altogether possible that these few samples of rhymed poetry were all written during the Islamic period in conscious imitation of Arabic poetry. In this case one would, however, have to conclude that rhyme—the more immediately tangible of the two essential components of Arabic poetry—was introduced into Persian earlier than quantitative metre. Clearly post-Islamic examples of rhymed, but not quantitative poetry include the couplet attributed, most probably wrongly, to Abū Ḥafṣ al-Sughdī, to which Lazard¹⁵ has contributed an important study, and perhaps a few other small fragments, of which, however, one cannot be certain whether they are in fact poems and not merely rhyming jingles.¹⁶ As for the so-called verse translation ¶ of the Qurʾān, a fragment of which was discovered in Qum,¹⁷ and which, according to its editor, represents ‘a bridge between accentual and quantitative poetry’, in the light of the unanimous insistence by Islamic tradition that the Qurʾān, despite its rhymed periods, and despite the fact that a good number of its āyahs can be scanned according to the rules of Arabic prosody, is not poetry (shiʿr), one must seriously doubt whether this Persian paraphrase was intended to be anything more than an imitation of the rhymed prose of the Arabic original.

According to an anecdote in the Tārīkh i Sīstān,¹⁸ the first person to compose a poem in (Neo-)Persian was Muḥammad b. Waṣīf,¹⁹ a secretary to the Saffarid ruler Yaʿqūb b. Laith. When the latter declared that he was unable to understand the Arabic verses that were being recited in his honour Muḥammad b. Waṣīf extemporised some laudatory verses in Persian. His example was then immediately followed by Bassām i Kūrd and Muḥammad b. Mukhallad. And thus Persian poetry was born. The story is, of course, not necessarily true and indeed the question of who was the ‘first Persian poet’ is essentially a futile one.²⁰ But the anecdote does certainly do justice to the circumstances, and to the approximate time (middle of the 3rd/9th century), in which Neo-Persian poetry came into being. When the court poets, who were accustomed to eulogising their masters in an already highly stylised type of Arabic poetry, found that they were no longer understood by the Iranian soldier-kings who had seized power on the Eastern fringe of the dār al-islām, they were compelled to switch to Persian and to perpetuate in that language the same forms and poetic traditions with which they were accustomed from Arabic. Although now working in Persian, they continued to produce poems of a distinctly Arabic type, employing mono-rhyme, quantitative metre as well as a perhaps somewhat simplified version of the same stylised imagery.

But to say that the early Persian poets borrowed the principles of quantitative metre from Arabic does not necessarily mean that the metres that they used were the same as those already in use in Arabic poetry. Their procedure ¶ was not quite the same as that of the Roman poets, who not only adopted the general poetic canons of the Greeks, but even forced their language into the same metres which the Greek tradition held appropriate for the several types of poetry: hexameter for heroic and didactic poems, the Lesbian metres for lyrics, iambics for tragic dialogue, etc. To be sure, some of the most common metres in Persian, such as mutaqārib muthamman maḥdhūf (or sālim): U--U--U--U-(-), hazaj musaddas maḥdhūf: U---U---U--, and ramal musaddas maḥdhūf: -U---U---U- occur also in basically the same form in Arabic, with the difference only that the Arabic mutaqārib,²¹ hazaj and ramal all permit in certain places the use of either a long or a short syllable (ancepites); the mutaqārib, for example, scans U-XU-XU-XU-(-); while in the corresponding Persian metres the places in question are always occupied by long syllables. These Persian metres thus make the impression of having been modelled on the ‘sound’, school-book versions of the corresponding Arabic metres, in which the long ancepites are regarded as standard, the short ancepites as a permitted variation of the standard pattern. On the other hand, many Persian metres, particularly those commonly used in lyric poetry, do not correspond to any Arabic metre, this despite the fact that the traditional Persian prosodic theory has given them elaborate Arabic names and attempted to ‘derive’ them from the standard Arabic metres with which they share a name. It is thus clear that the pioneers of Persian poetry, besides borrowing, or rather adapting, some of the Arabic metres, also developed a number of new, purely Persian metres of an Arabic (i.e. quantitative) type. To these the prosodists later assigned more or less artificial Arabic names.

In his important book The Persian metres²² L.P. Elwell-Sutton has shown the inadequacy of the traditional Arabic-based analysis of Persian metres, refuted (I should think for good) the notion that the latter can be derived totally from Arabic models and laid the foundation for a new approach to the formal analysis of Persian poetry on the basis of the prosodic patterns actually occurring in Persian verse. However, Elwell-Sutton went a step further and claimed that the Persian system of quantitative metre has in fact nothing to do with Arabic, but continues the formal traditions of pre-Islamic Persian poetry. This claim ¶ is, however, totally unsubstantiated. Neither Elwell-Sutton nor anyone else has succeeded in analysing Old or Middle-Iranian poetry (of which a considerable amount has survived) along quantitative lines, and as long as such an analysis has not succeeded we cannot but assume that that poetry was not quantitative. In fact, as has already been mentioned, a very strong case for an accentual basis of Middle-Iranian poetry has been made by such experts as Henning and Boyce. It is most regrettable that Elwell-Sutton, by referring to antiquated studies by Iranicists like Benveniste, Nyberg, and Christensen, or non-Iranicists like Marr,²³ must inevitably induce non-specialist readers to think that the study of pre-Islamic Iranian poetry is a field where anything goes. In fact it is one where there is now a fairly large degree of consensus among competent judges.

The question of why the Persian poets of the Islamic period invented the particular non-Arabic metres that they did is one about which one can only hazard a guess. It would seem, however, most probable that educated Persian Muslims of the first centuries after the hijrah, schooled as they were in Arabic poetry and, perhaps more importantly, in tajwīd, the science of Qurʾānic recitation, with its painfully exact measurement of the length of every syllable, must have become aware of the varying length of the syllables in their own language, as well as of the metrical patterns typical of it. When they began to compose poetry of an Arabic type in their own language they obviously felt it imperative to maintain a consistent pattern of long and short syllables in each verse. It was not necessary, however, to use the actual patterns occurring in Arabic poems. It was left to the metricians of later times to analyse these Persian patterns and to force them more or less violently into the scheme devised by the Arabic prosodists.

If we ignore stanzaic poetry, which in all periods of Persian literature has played only a marginal role,²⁴ we can divide Persian poems into those which have monorhyme (i.e. the same rhyme occurs at the conclusion of every verse from the beginning to the end of the poem), on the one hand, and mathnawīs (i.e. those consisting of rhymed couplets) on the other.²⁵ Poems ¶ with monorhyme, which we can conveniently designate as ‘lyric’ verse, range from two-line epigrams to odes of more than a hundred verses, though really long odes are very much less common in Persian than in Arabic. In the early periods lyric poems are most frequently panegyrics (poems flattering a king or some other patron), though we also find elegiac-didactic poems (often lamenting old age, or expressing pessimistic sentiments), erotic and bacchic poems, lampoons (generally directed against rival poets or tight-fisted ex-patrons) and assorted facetiae. Religious pieces are relatively rare before the time of Sanāʾī (first half of the 6th/12th century), though later they become the dominant type of lyric poetry. All of these genres are well-known in Arabic and the stylised structural and rhetorical devices of Arabic lyrics are imitated freely by the Persian authors.

Poems consisting of rhymed couplets are generally long to very long (the Shāh-nāmah has about 60 000 verses) and of either narrative or didactic content. Indeed the distinction between narrative and didactic mathnawīs is not always clear-cut; Asadī’s Karshāsp-nāmah, for example, though fundamentally a story-poem, indulges in long paraenetic excursions, as do many later epics. Although long instructive and narrative poems in rhymed couplets are by no means unknown in Arabic, they are considerably less common there than in Persian and have certainly never enjoyed the same status as the Persian mathnawī. Moreover, in Arabic this sort of poetry is restricted to a single metre (rajaz), while writers of Persian mathnawīs have a greater choice of metres, though even they use only half a dozen metres with any frequency. The subject-matter of Persian narrative poems is in many cases taken from the legendary and semi-legendary traditions of pre-Islamic Iran. This is the case with Firdausī’s Shāh-nāmah, Asadī’s Karshāsp-nāmah and the other heroic poems belonging to the Persian ‘epic cycle’,²⁶ but also such romantic epics as Gurgānī’s Wīs u Rāmīn or Niẓāmī’s Khusrau-Shīrīn, as well as overtly Zoroastrian works like Kai-Kāʾōs’s Maulūd i Zartusht. But there are also poems based on Arabic, Islamic and Islamicised Biblical traditions such as the various versions of Yūsuf u Zulaikhā, ʿAiyūqī’s Warqah u Gulshāh or Niẓāmī’s Lailē-Majnūn. Others again can be traced to Hellenistic sources, e.g. ʿUnṣurī’s Wāmiq u ʿAdhrā or the versions of the legend of Alexander by Niẓāmī and others. Three of the very earliest Persian mathnawīs of which we have any knowledge, namely Rōdakī’s Kalīlah u ¶ Dimnah and Sindbād-nāmah, and the anonymous, but roughly contemporary Bilauhar u Būdhāsaf,²⁷ all retell in Persian the same (ultimately Sasanian or Sasanianised Indian) stories which Abān al-Lāḥiqī had put into Arabic rhymed couplets²⁸ more than a century earlier. All three of these Persian poems are in ramal metre, which, as can be seen from the table, is based on the same recurring pattern as rajaz, the canonical metre of Arabic couplets. And the same pattern underlies also the hazaj metre employed in the earliest poetical version of the Shāh-nāmah, that by Masʿūdī al-Marwazī,²⁹ and in other early mathnawīs, whether narrative (Wīs u Rāmīn) or didactic (e.g. Maisarī’s Dānish-nāmah) and well as in many long poems of later authors (e.g. Niẓāmī’s Khusrau-Shīrīn and its many imitations or Rūmī’s Mathnawī i maʿnawī). It is thus quite clear that the narrative and didactic poetry of the Persians, like their lyrical verse, had Arabic antecedents both for its content and for its general form.

rajaz: --U---U---U- etc.

ramal: -U---U---U-- etc.

hazaj: U---U---U--- etc.

One of the striking features of early Persian narrative poems is that their authors repeatedly and insistently tell us that their poems are based on an ‘old book’. In other words, they are versifications of pre-existing written narratives. In at least one case the source used by the poet is actually extant (the Middle-Persian Zand i Wahman-Yasht used by Kai-Kāʾōs³⁰) and in some others it can at least be identified (e.g. Abū Muʾaiyad’s Kitāb i Karshāsp as the probable source of Asadī’s epic). In general these sources can be assumed to have been in Persian prose and to have been translated either from an Arabic or a Middle-Persian original, though in a few cases (Maulūd i Zartusht and Wīs u Rāmīn) the poet appears to have worked directly on the basis of a source in Middle Persian.³¹

Despite the insistence by the authors of these narratives that they are merely retelling what they found in a ‘book’, attempts have occasionally been made to view early Persian poetry in the light of the well-known theory of ‘oral ¶ poetry’,³² a theory which has had a very strong influence particularly on the Anglo-Saxon school of Homeric studies, but which has also been applied with interesting results to such fields as pre-Islamic Arabic poetry. In the case of Persian, this theory would imply the existence of an uninterrupted tradition of poetry handed down from generation to generation from remote Iranian antiquity all the way to the professional Shāh-nāmah-reciters of the present. Support for this conception might be sought in the fact that in Firdausī’s Shāhnāmah, alongside the many passages where the poet speaks of the ‘old book’ that he has put into verse, there are also a number of places where he states, or implies, that he has ‘heard’ the story he is about to tell from an ‘old dihqān’ or the like.³³ But it is much more likely that in all the passages of this sort the poet is merely repeating, in verse, the statement by his written source that it has derived its information from the person in question. Thus, when at the beginning of the story of Burzōy³⁴ the poet invites us to listen to the words of Shādān, the son of Burzīn, the casual reader could be forgiven for thinking that Firdausī actually heard Shādān tell this story. In fact, we know from the ‘older preface’ to the Shāh-nāmah³⁵ that Shādān, son of Burzīn, was one of the ‘four men’ who collaborated in the compilation of the prose Shāh-nāmah that was written for Abū Manṣūr b. ʿAbd al-Razzāq in Muḥarram 346/957, one of Firdausī’s written sources.³⁶ Firdausī’s reference to Shādān is thus clearly lifted directly from his source. Further on in the text, at the beginning of the story of Hurmuzd,³⁷ the poet speaks quite vividly of a meeting which he had with Mākh, the margrave of Herat, a wise old man, whom Firdausī asks for information about the reign of king Hurmuzd and whose reply the poet then proceeds to quote. But this is manifestly a poetic fiction. Mākh the marzbān of Herat is clearly identical with another of the ‘four men’ whose name appears in the manuscripts of the ‘older preface’ as Shāj of Herat ماخ is either a misreading of شاج, or vice versa). The same is apparently also the case with the story of the invention of the game of chess which Firdausī tells on the authority of ‘old Shāhōy’,³⁸ evidently yet another of the ‘four men’, whose name occurs in the ‘older preface’ as Māhōy (this, like the previous example, must be seen in the ¶ light of the close similarity of initial m- and s/sh- in early Persian handwriting). It is, of course, not out of the question that the young Firdausī might actually have met one or the other of the ‘four men’, but it is rather unlikely that he should have known three of the four.

Further evidence for the dependence of the Persian epic tradition on written sources, rather than on a living oral tradition can be seen in the fact that a good number of the proper names that figure in that tradition appear in a form that can only be explained in terms of the misreading of written sources. Thus the name of Ṭahmūrath (or -rat) clearly results from the mispointing in Arabic script of *Ṭahmūrab, for Middle-Persian tʾḥmwlp and Avestan Taxma- Urupi-.³⁹ Similarly the name given to Farēdūn’s evil son Tūr⁴⁰ is evidently a mispointing (again conceivable only in Arabic script) for Tūz,⁴¹ representing Middle-Persian twc.⁴² Or Isfandyār,⁴³ a corruption in Arabic script for Isfandyād (or Isp-),⁴⁴ for Middle-Persian spndyʾt or spnddʾt,⁴⁵ for Avestan Spəntōδāta-.⁴⁶ Or Nastūr, a mispointing of Bastūr, Middle-Persian bstwl, Avestan Bastauuairī.⁴⁷ For G/Karshāsp, from Karsāsp, see below, p. 30, footnote no. 52. The question of whether all these spellings were in fact already used by Firdausī and are not merely the result of later scribal corruptions must, for the adherents of the theory of oral poetry, be irrelevant, since the prerequisite of this theory is precisely the assumption of an uninterrupted oral tradition from antiquity to the present day. Such a tradition ought not to be able to be led astray by scribal errors, whether before or after the time of Firdausī, who is but one link in an unbroken chain of oral poets.

It is quite clear that a theory of oral poetry will only work if it is possible to posit a continuing formal tradition of versification. In the case of the Homeric epics, for example, it is assumed that the stories of the Trojan War and of the adventures of Odysseus were retold by generation after generation of bards, all ¶ using the same hexametre verse, the same (or much the same) somewhat artificial poetic language (which was not identical with any of the spoken Greek dialects), the same set similes and standard epithets, etc., until the time when one particular version of the poems was set down in writing. Of course, we cannot prove that this was actually the case, but it is not an implausible hypothesis. But Firdausī and his contemporaries did not have this kind of tradition behind them. They were pioneers. Persian poetry with rhyme and quantitative metre was, as we have seen, only a bit more than a century old. This innovation cut them off from the old poetic tradition. When they told the same stories as the pre-Islamic minstrels, as is evidently the case with Gurgānī’s Wīs u Rāmīn,⁴⁸ their link with their predecessors was through books, not through a living tradition.

Our conclusion can thus only be that the Shāh-nāmah, as it was written by Firdausī, was not oral poetry, but book-literature. However, almost as soon as it was written down, it most certainly did turn into oral poetry on the tongues of the rhapsodists, who developed and elaborated the epic orally and have continued to do so to the present. The tremendous degree of disagreement already between the oldest manuscripts of the poem cannot be explained purely in terms of the carelessness and unscrupulousness of generations of scribes. It is quite clear that from a very early date the scribal and oral textual traditions have constantly influenced one another. But this is an oral tradition which does not (as is assumed to have been the case with the Homeric poems) culminate and end with a book. In Iran the book is the point of departure.

Addendum (December 2000): This chapter has been reprinted as it appeared in the first edition (1992), with correction of a few mistakes and some additions to the bibliography.

The origin of the Persian metrical system, and its relationship with the Arabic, have since been the subject of several articles (all without reference to our discussion) in the volume Arabic prosody and its applications in Muslim poetry, ed. L. Johanson and B. Utas (=Swedish Research Institute in Istanbul, Transactions, Vol. 5), Uppsala 1994; see in particular: pp. 35–43 (J.T.P. de Bruijn, ‘The individuality of the Persian metre khafîf’); pp. 45–59 (G. Doerfer, ‘Gedanken zur Entstehung des rubāʿī’); pp. 81–90 (G. Lazard, ‘Le mètre épique baloutchi et les origines du motaqâreb’); pp. 129–41 (B. Utas, ‘Arabic and Iranian elements in New Persian prosody’).

O. Davidson’s essay (mentioned above, p. 9, fn.) has been republished (with slight reworking) as the first half of her book Poet and hero in the Persian Book of Kings, Ithaca and London 1994, and a similar line of reasoning has now been ¶ pursued by D. Davis in his article ‘The problem of Ferdowsî’s sources’, jaos 116, 1996, pp. 48–57, again both without reference to the arguments in the present book. Davidson’s book has been criticised by (among others) M. Omidsalar in jaos 116, 1996, pp. 235–42, G. Herrmann in Central Asiatic Journal xli, 1997, pp. 127–8, and by me in jras 1998, pp. 269–70.

In the addenda to the first edition (pp. 612–4) I drew attention to the fact that a further sample of (apparently) pre-Islamic narrative poetry is the fragment quoted in the Vatican manuscript of Asadī’s lf (ed. Horn p. 34; Salimi p. 274; omitted in Iqbāl’s edition!) from ‘the book of Pīrān i Wēsah in the Pahlawī language’.⁴⁹ The Ms. has:

Nöldeke (cited in Horn’s footnote) recognised the words tundur, aβrōčand and sāčāδ in the second ‘half-verse’ and Henning (in the hand-written notes in his copy of Horn, kindly lent to me by N. Sims-Williams) emended the first four words as given below. Standing on the shoulders of these illustrious predecessors we might propose reading the whole thing as: azd-am kard, šahryār, kū-t hīr wasnāδ *tundur aβrōjīd u tum sājād which could mean something like: ‘I have made it known, oh king, that it is for your sake that the thunder was ignited⁵⁰ and the darkness brought forth’. Pīrān, son of Wēsah, figures in the Shāh-nāmah as a courtier of Afrāsyāb, but it is difficult to place the verses in the story. The language of the fragment is like that of Draxt ī Āsūrīg and Ayādgār ī Zarērān in that it is basically Parthian with an admixture of Persian,⁵¹ a Partho-Persian poetic koine, and thus comparable to the language of the Homeric poems (mainly Ionic, but with an admixture of other dialects). Both lines have nine syllables and (it seems) four stresses each, but whether or not the apparent ‘rhyme’ between wasnāδ and sāžād is intentional can hardly be judged from a single pair of lines.

next chapter: Part 1